Face detected with its landmarks
Real face

Straight angle, no filter

Adult

Minor probability: 1%

Perfect quality

No obstruction

Face Detection that sees past masks, phones and filters

Detect faces in images and videos, including hidden, disguised and edited ones. Then go further: tell real people from artwork, measure obstruction and quality, spot filters and detect minors.

Have questions? Talk to sales

Powering some of the most innovative platforms worldwide

ADVERSARIAL FACE DETECTION

Find the faces other detectors miss

People hide their faces, on purpose or not. Our model is trained on the hard cases: faces behind the phone in mirror selfies, under masks, hoods and scarves, pixelated or blurred, covered by filters and edits, turned away from the camera or cut by the frame.

Mirror selfiesMasksHoods & scarvesPixelationBlurFilters & overlaysStrong anglesTight crops

Each face comes with its bounding box and five landmarks: eyes, nose tip and mouth corners, estimated even when hidden. Works on images and videos.

Explore Face Detection
Faces partly hidden, cut, at an angle or blurred

ADVANCED FACE ANALYSIS

Real person, or just a face?

Separate real human faces from drawings, paintings, 3D renders, statues, dolls, masks and mannequins. Artificial faces are reported separately, so they never pass for a person.

Real vs artificial faces
Artificial face

Statue

Statue detected as an artificial face

Measure how much of the face is hidden

Obstruction is graded on six levels, from none to complete, based on which key features (eyes, nose, mouth) can still be seen. Properly worn sunglasses are reported on their own.

Obstruction levels
None
None
Light
Light
Medium
Medium
Heavy
Heavy
Extreme
Extreme
Complete
Complete

Know if the face is clear enough

Face quality is graded on four levels, from perfect to low, from lighting, noise, blur, pixelation and distortion. The face angle is returned too: straight, side or back.

Quality levels
Perfect
Perfect
High
High
Medium
Medium
Low
Low

Spot filters and face edits

Detect visible face filters, whether they cover the face (fake glasses, masks, stickers, text) or distort it (warped features, cartoon effects). Subtle beauty filters and global color edits are ignored.

Filter detection
Faces with distorting and obstructing filters

Is the person actually there?

Tell faces of people physically present in the scene, mirror reflections included, from faces shown on phones, screens, posters and magazine covers, or pasted onto the image.

In-scene detection
Not in scene

Face on a screen

Face shown on a phone screen

AGE ANALYSIS

Detect minors and estimate age

For every real face, get the probability that the person is under 18, to keep minors off adult and age-restricted platforms, or to blur their faces. Finer age groups, from toddlers to seniors, are available for advanced needs.

Explore Minor Detection Ask about age groups
Minor

Under 18

Child detected as a minor

Validate profile pictures and selfies

Combine face analysis with nudity, AI-generated image and liveness detection in a single API call. Require one clear, real, unobstructed face, keep minors out and stop fake accounts at signup.

Face Liveness Detection Blur faces
Valid
Valid
Face hidden
Face hidden
Back angle
Back angle
Low quality
Low quality
Drawing
Drawing
Several faces
Several faces